Fine-tuning four open-source LLMs on LLM-generated Socratic guidance data changes tutor behavior, reducing answer disclosure on some models, but gains are inconsistent and are measured by an LLM judge rather than human learners.
No expert solution available for this question
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
CoDAE: Adapting Large Language Models for Education via Chain-of-Thought Data Augmentation
Fine-tuning four open-source LLMs on LLM-generated Socratic guidance data changes tutor behavior, reducing answer disclosure on some models, but gains are inconsistent and are measured by an LLM judge rather than human learners.